modulator NS11394 with diazepam, zolpidem, bretazenil and gaboxadol in rat

Transcription

1 JPET Fast This Forward. article has not Published been copyedited on and September formatted. The 12, final 2008 version as may DOI: /jpet differ from this version. Comparison of the novel subtype selective GABA A receptor positive allosteric modulator NS11394 with diazepam, zolpidem, bretazenil and gaboxadol in rat models of inflammatory and neuropathic pain Munro G, Lopez-Garcia JA, Rivera-Arconada I, Erichsen HK, Nielsen EØ, Larsen JS, Ahring PK, Mirza NR * NeuroSearch A/S, 93 Pederstrupvej, DK-2750, Ballerup, Denmark (G.M., H.K.E., E.Ø.N, P.K.A, J.S.L., N.R.M.); Department of Physiology, University of Alcalá, Madrid, E Spain (I.R.-A., J.A.L.-G.) 1 Copyright 2008 by the American Society for Pharmacology and Experimental Therapeutics.

2 Running title: Antinociceptive actions of NS11394 in rat pain models *Corresponding author: Dr Naheed Mirza Department of Pharmacology NeuroSearch A/S 93 Pederstrupvej DK-2750, Ballerup, Denmark - max@neurosearch.dk Telephone ; Fax; Number of text pages: 46 Number of tables: 2 Number of figures: 8 Number of references: 38 Number of words in Abstract: 250 Number of words in Introduction: 749 Number of words in Discussion: 1549 Abbreviations: ACSF, artificial cerebrospinal fluid; CCI, chronic constriction injury; CFA, complete Freund s adjuvant; CNS, central nervous system; DR-VRr, dorsal rootventral root reflex; GAD, glutamic acid decarboxylase; GABA, gamma-amino-butyric acid; IPSP, inhibitory postsynaptic potential; MA, mechanical allodynia; MED, minimum effective dose; MH mechanical hyperalgesia; MPE, maximum possible effect; SNI, spared nerve injury; TF, tail flick; TRPA1, transient receptor potential A1; TRPV1, transient receptor potential V1; WB, weight bearing 2

3 Recommended section assignment: Neuropharmacology 3

4 Abstract Spinal administration of GABA A receptor modulators such as the benzodiazepine drug diazepam partially alleviates neuropathic hypersensitivity that manifests as spontaneous pain, allodynia and hyperalgesia. However, benzodiazepines are hindered by sedative impairments and other side-effect issues occurring mainly as a consequence of binding to GABA A receptors containing the α1 subunit. Here, we report on the novel subtype selective GABA A receptor positive modulator NS11394 [3 -[5-(1-hydroxy-1-methylethyl)-benzoimidazol-1-yl]-biphenyl-2-carbonitrile] which possesses a functional efficacy selectivity profile of α5 > α3 > α2 > α1 at GABA A α subunit-containing receptors. Oral administration of NS11394 (1-30 mg/kg) to rats attenuated spontaneous nociceptive behaviours in response to hindpaw injection of formalin and capsaicin, effects that were blocked by the benzodiazepine site antagonist flumazenil. Ongoing inflammatory nociception observed as hindpaw weight-bearing deficits after Freunds adjuvant injection was also completely reversed by NS Similarly, hindpaw mechanical allodynia was fully reversed by NS11394 in two rat models of peripheral neuropathic pain. Importantly, NS11394-mediated antinociception occurred at doses fold lower than those inducing minor sedative or ataxic impairments. In contrast, putative antinociception associated with administration of either diazepam, zolpidem or gaboxadol only occurred at doses producing intolerable side-effects, whereas bretazenil was completely inactive despite minor influences on motoric function. In electrophysiological studies, NS11394 selectively attenuated spinal nociceptive reflexes and C-fibre mediated wind-up in vitro pointing to involvement of a spinal site of action. The robust therapeutic window seen with NS11394 in animals suggests that compounds 4

5 with this in vitro selectivity profile could have potential benefit in clinical treatment of pain in humans. 5

6 Introduction Within the mammalian spinal cord, GABA is the principal inhibitory transmitter and is localized both pre-synaptically in primary afferents and post-synaptically in dorsal horn interneurones (Malcangio and Bowery, 1996). Immunohistochemical studies suggest all major classes of GABA A receptors, e.g. those containing α1, α2, α3 and α5 subunits (hereafter referred to as GABA A -αx), are variously expressed within the spinal dorsal horn, although GABA A -α2 receptors predominate within superficial layers which receive nociceptive input from primary afferents (Bohlhalter et al., 1996; Knabl et al., 2008). Loss of GABA function contributes to central hyperexcitability associated with tissue injury, a process referred to as spinal disinhibition. After partial or complete nerve injury, primary afferent evoked IPSPs in dorsal horn neurones are substantially reduced in frequency, magnitude and duration, likely as a consequence of reduced GABA release (Moore et al., 2002). Nerve injury also modulates levels of GABA, the GABA synthesizing enzyme GAD and GABA A receptor expression within the spinal dorsal horn (Castro-Lopes et al., 1993; Castro-Lopes et al., 1995; Moore et al., 2002). At the behavioural level, GABA A receptor agonists such as muscimol and positive allosteric modulators such as diazepam attenuate nociceptive transmission in animal models of persistent and neuropathic pain associated with central sensitization (Kaneko and Hammond, 1997; Hwang and Yaksh, 1997; Malan et al., 2002). Why then, have encouraging preclinical observations on the functional importance of GABA A receptors in the mammalian spinal cord failed to translate into successful clinical treatments for pain? The majority of GABA A receptors are sensitive to allosteric modulation by benzodiazepine drugs and typically contain the α subunits α1, α2, α3 or 6

7 α5, together with a β subunit and a γ2 subunit in a 2:2:1 stoichiometry (Sieghart 1995). Pharmacological studies, recently combined with the use of genetically engineered mice, have led to a general consensus that GABA A -α1 receptors mediate sedation, GABA A -α2/gaba A -α3 receptors are involved in anxiety, whilst GABA A -α5 receptors are relevant to memory function (Squires et al., 1979; Griebel et al., 1999a; 1999b; Paronis et al., 2001; Rudolph and Möhler, 2006). Recently, using a combination of molecular and pharmacological techniques Knabl et al. reported that GABA A -α2/ GABA A -α3 receptors are the principal contributors to spinal disinhibition occurring after injury (Knabl et al., 2008). Importantly, the prototype modulator tested (L- 838,417), was analgesic in a range of rat models of persistent pain, and was devoid of sedative or motor impairing qualities; albeit no therapeutic window was provided. Thus, development of subtype selective GABA A receptor modulators designed to normalize spinal inhibition after injury might offer a novel mechanistic approach for treating neuropathic pain. To investigate further, we decided to test for antinociceptive actions of the subtype selective GABA A receptor positive modulator NS11394 (Figure 1) in a range of animal models of chronic pain. NS11394 binds potently and non-selectively (Ki, ~0.5 nm) to GABA A receptors containing either α1, α2, α3 or α5 subunits combined with β2 and γ2s subunits, whereas affinity at GABA A -α4 or GABA A -α6 containing receptors is fold lower. Despite this lack of affinity selectivity, in electrophysiological studies NS11394 differentially modulates these receptors, such that it possesses a functional selectivity profile of α5 > α3 > α2 > α1. This in vitro profile translates to potent ( mg/kg, p.o.) anxiolytic-like effects in rodent shock/non-shock based 7

8 models of anxiety, likely attributable to its efficacy at GABA A -α3 and/or GABA A -α2 receptors (Mirza et al., 2008). By contrast, at doses fold higher NS11394 minimally impairs motor performance, likely attributable to low efficacy at GABA A -α1 receptors and poor affinity for GABA A -α4 and GABA A -α6 receptors. To obtain a comprehensive picture of the role of GABA A receptors in mediating painlike behaviours after injury, we compared NS11394 to other known allosteric modulators: the full non-selective modulator diazepam, the partial non-selective modulator bretazenil, and the GABA A -α1 selective full modulator zolpidem. Furthermore, we included the GABA site partial/full agonist gaboxadol to compare the actions of modulators to that of direct receptor activation (Ebert et al., 2006). This comparative pharmacological approach allows us to delineate pharmacologically which GABA A receptors might be relevant to pain, and therefore complements recent findings in transgenic mice (Knabl et al., 2008). However, the approach here is necessary per se to understand pharmacological nuances not easily discernable using a transgenic mouse approach: e.g., level of functional efficacy in addition to selectivity necessary for invivo efficacy. We also compare the motor side-effects of the various GABA A compounds thereby deriving therapeutic indices, a fundamental cornerstone in the development of novel therapeutics. Finally, we report on mechanistic studies to determine the effects of NS11394 on the spinal nociceptive circuits which mediate withdrawal reflexes. 8

9 Methods Animals Adult male Sprague-Dawley rats (Harlan Scandinavia, Alleroed, Denmark) were used except where stated. They were housed in Macrolon III cages (20 x 14 x 18 cm or 20 x 40 x18 cm; in groups of 2-5 per cage according to weight) containing wood-chip bedding material (3 x 1 x 4 mm), and in a temperature-controlled environment with a light-dark cycle of 13:11 h (lights on at h and off at h). Food (Altromin ) and water were available ad libitum. The animals were allowed to habituate to the housing facilities for at least one week prior to surgery or behavioural testing. Neuropathic animals were subsequently housed on soft bedding material. All behavioural experiments were performed according to the Ethical Guidelines of the International Association for the Study of Pain (Zimmermann, 1983) and licensed by the Animals Experiments Inspectorate, The Danish Ministry of Justice. Behavioural tests Explorative motility and Rotarod test Non-habituated male rats (body weight g) were administered test drug or vehicle and returned to their home cage for min depending on the drug administered. They were then placed individually in transparent cages (30 x 20 x 25 cm, TSE Systems, Bad Homburg, Germany) equipped with 12 infrared sensors (6 x 2). Locomotor activity, measured as distance travelled (m), was monitored automatically in the chambers via the interruption of two consecutive infrared sensors. Interruptions were detected by a control unit and recorded via a computer running ActiMot software (TSE Systems). Raw data obtained via 3 min sampling intervals was summed for the 30 9

10 min duration of the experiment, and expressed as a % of the vehicle response according to the equation, (1) % Vehicle = (Post-treatment value/vehicle value) 100. In normal, uninjured rats (body weight g) the effects of GABA A receptor modulators on activity-induced motor function was evaluated using an accelerating rotarod (Ugo Basile, Comero, Italy). The rotarod (6 cm diameter) speed was increased from 3-30 rpm over a 180 s period, with the minimum time possible to spend on the rod designated as 0 s and the maximum cut-off time set at 180 s. Rats received two training trials (separated by 3-4 h) on two separate days prior to drug testing for acclimatisation purposes. On the day of drug testing, a baseline response was obtained, and rats subsequently administered drug or vehicle with the time course of motor performance tested at 30 min after injection for GABA A receptor modulators and agonists and 60 min after injection of gabapentin. Raw data was subsequently expressed as a % of the corresponding baseline response according to the equation, (2) % Baseline = (Post treatment value/baseline value) 100. Hot plate For the hot plate test, non-injured rats (body weight g) were placed individually on a hot plate (Ugo Basile) maintained at a constant pre-set temperature of 52.5 C with a cut-off time of 40 s. The baseline latency response (s) for the rat to lift either hindpaw induced by the thermal stimulus was then measured, at which point the animal was 10

11 immediately removed from the hot plate by the investigator to prevent any tissue damage to the hindpaws. Animals were then administered drug or vehicle and posttreatment latency responses were determined at min after injection depending on the drug administered, enabling the change in latency response (s) to be calculated. For all experiments involving hot plate measurements, raw data obtained at the time points indicated was converted to a maximal possible effect value according to the equation, (3) %MPE = (Post-treatment value - Pre-treatment value) 100/(Ceiling value of assay - Pre-treatment value). Formalin test and capsaicin-induced sensitization Assessment of formalin-induced and capsaicin-induced flinching behaviour in normal, uninjured rats (body weight g) was made with the use of an Automated Nociception Analyzer (University of California, San Diego, CA). Briefly, this involved placing a small C-shaped metal band (10 mm wide x 27 mm long) around the hindpaw of the rat to be tested. Each rat (four rats were included in each testing session) was administered drug or vehicle according to the experimental paradigm being followed, and then placed in a cylindrical acrylic observation chamber (diameter 30.5 cm x height 15 cm). For formalin experiments, individual rats were then gently restrained and formalin (5% in saline, 50 μl, s.c.) was injected into the dorsal surface of the hindpaw using a 27G needle. For capsaicin-induced sensitization experiments, individual rats were gently restrained and capsaicin (10 μg in 10 μl 10% Tween 80, s.c.) was injected into the plantar surface of the hindpaw using a 0.3 ml insulin syringe with a 29G needle 11

12 (Terumo Europe, Belgium). All rats were immediately returned to their separate observation chambers, each of which was situated upon an enclosed detection device consisting of two electromagnetic coils designed to produce an electromagnetic field in which movement of the metal band could be detected. The analogue signal was then digitised and a software algorithm applied to enable discrimination of flinching behaviour from other paw movements. A sampling interval of 1 min was used. On the basis of the resulting response patterns, for formalin experiments three phases of nociceptive behaviour were identified and scored; first phase (0-5 min), interphase (6-15 min) and second phase (16-40 min), (Munro et al., 2008). For capsaicin-induced sensitization experiments two phases of nociceptive behaviour were identified and scored; first phase (0-5 min) and second phase (6-30 min). For both formalin and capsaicin experiments raw data from the 1 min sampling intervals was summed for each phase to obtain the total number of flinches occurring during that phase. This value was then expressed as a % of the vehicle response according to the equation, (4) % Vehicle = (Post-treatment value)/(vehicle value) 100. Complete Freunds adjuvant-induced inflammatory nociception Rats (body weight g) were given a s.c. injection of CFA (50% in saline, 100 μl, Sigma) into the plantar surface of the hindpaw under brief halothane anaesthesia. Nociceptive behaviours were routinely assessed for 2-3 days prior to, and 24 h following CFA injection. Changes in hindpaw weight bearing were assessed using an Incapacitance tester (Linton Instrumentation, U.K.), which incorporates a dual channel scale used to separately assess the weight distributed to each hindpaw of the rat. 12

13 Normally, uninjured rats distribute their weight evenly between the two hindpaws (50:50). However, after tissue injury the rat preferentially favours the non-injured hindpaw, such that the weight bearing difference can be used as an index of spontaneous nociception. A rat was placed in the supplied Perspex chamber which is designed so that each hindpaw must be placed on separate transducer pads. The testing duration was set to 5 s and the digital read out for each hindpaw was taken as the distributed body weight on each paw (g). Three readings were obtained to ensure that consistent responses were measured. These were averaged for each hindpaw and the weight bearing difference calculated as the difference between the two hindpaws. Tail flick responses were also measured in the same CFA rats. A radiant heat source (Ugo Basile) was focused on the underside of the rat s tail 3 cm from its distal end, with the apparatus calibrated to give a tail flick latency of approximately 4-6 s (cut off time = 15 s ). This enabled increases or decreases in tail flick latency to be measured to the nearest 0.1 s. Two baseline measurements (2 measurements each separated by 5 min) were made prior to CFA injection, to familiarise the rats with the testing procedure. Two further baseline latency measurements were obtained on the day of drug testing to ensure consistent reflex responses were present. Animals were then administered drug or vehicle according to the experimental paradigm, with weight bearing responses determined at 30, 60 or 120 min post-injection and tail flick latency responses determined at min depending on the drug administered. Weight bearing differences after drug treatment were expressed as a % of the corresponding baseline response according to the equation, 13

14 (5) % Baseline = (Post-treatment value)/(baseline value) 100. Tail flick latencies were converted to a maximal possible effect value according to the equation, (6) %MPE = (Post-treatment value - Pre-treatment value) 100/(ceiling value of assay - Pre-treatment value). Peripheral nerve injury A chronic constriction injury (CCI) or spared nerve injury (SNI) was performed in rats (body weight g at the time of surgery) as described previously (Bennett and Xie, 1988; Decosterd and Woolf, 2000). Anaesthesia was induced and maintained by 2% isoflurane (Baxter A/S, Alleroed, Denmark), combined with oxygen (30%) and nitrous oxide (68%). For CCI rats, the sciatic nerve was exposed at the mid-thigh level proximal to the sciatic trifurcation. Four chromic gut ligatures (4/0) (Ethicon, New Brunswick, NJ) were tied loosely around the nerve, 1-2 mm apart, such that the vascular supply was not overtly compromised. For SNI rats, the skin of the lateral left thigh was incised and the cranial and caudal parts of the biceps femoris muscle were separated and held apart by a retractor to expose the sciatic nerve and its three terminal branches: the sural, common peroneal and tibial nerves. The tibial and common peroneal nerves were tightly ligated with 4/0 silk and 2-3 mm of the nerve distal to the ligation was removed. Any stretching or contact with the intact sural nerve was avoided. For both CCI and SNI rats, the overlying muscle was closed in layers with 4/0 synthetic absorbable surgical suture. The skin was closed and sutured with 4/0 silk thread. 14

15 Behavioural testing of nerve-injured rats Nerve-injured rats were routinely tested for the presence of hindpaw mechanical hypersensitivity according to previously described methods (Munro et al., 2008). Prior to testing individual rats were removed from their home cage and allowed to habituate for 15 min in an openly ventilated 15 x 20 cm white Plexiglass testing cage, placed upon an elevated metal grid allowing access to the plantar surface of the injured hindpaw. The presence of mechanical allodynia was assessed using a series of calibrated von Frey hairs (lower limit=0.06 and upper limit=13.5 g, Stoelting Co, Wood Dale IL), which were applied to the plantar surface of the hindpaw (lateral aspect in SNI rats since this is the area of the paw innervated by the intact sural nerve) with increasing force until an individual filament used just started to bend. The filament was applied for a period of 1-2 s and was repeated 5 times at 1-2 s intervals. The filament that induced a reflex paw withdrawal in 3 out of 5 applications was considered to represent the threshold level for a mechanical allodynic response to occur. The presence of mechanical hyperalgesia was determined in SNI rats by pressing the plantar surface of the hindpaw with the point of a safety pin, at an intensity sufficient to produce a reflex withdrawal response in normal unoperated animals, but at an intensity which was insufficient to penetrate the skin. A cut-off time of 15 s was applied to long withdrawals often seen for the nerve-injured paw. For both von Frey hair and pin prick measurements raw data values obtained from min after drug administration were expressed as a %MPE response according to equation (6) described above. Only those animals showing distinct neuropathic behaviours from between days post-surgery were included in drug testing experiments. During an individual testing 15

16 session which used 6-8 nerve-injured rats nociceptive measurements were performed with the observer blinded to treatment. By strictly adhering to a minimum 2-3 day drug washout period between experiments, an escalating dose crossover paradigm could be used. Thus, although the majority of SNI and CCI rats typically received more than one drug treatment, no animal received more than (i) one injection of vehicle plus injection of the same drug at two different doses (ii) one injection of vehicle plus two injections of 2 mechanistically distinct drugs. In vitro electrophysiology The isolated rat spinal cord preparation The preparation and in vitro maintenance of the spinal cord followed procedures described previously (Hedo and Lopez-Garcia, 2001). Briefly, Wistar rat pups (7 11 days old) were anaesthetised with urethane (2 g/kg, i.p.) and their spinal cords extracted following a dorsal laminectomy. With the cord in cold artificial cerebrospinal fluid (ACSF), the outer meninges were removed, the cord was hemisected and then pinned down to a Sylgard based recording chamber with the medial side upwards. The preparation was maintained with oxygenated (95% O 2, 5% CO 2 ) ACSF at room temperature (23 ± 1 C). Flow rate was 5 ± 2 ml/min. The composition of the ACSF was (in mm) NaCl (128); KCl (1.9); KH 2 PO 4 (1.2); MgSO 4 (1.3); CaCl 2 (2.4); NaHCO 3 (26); glucose (10); (ph 7.4). All drugs tested in vitro were dissolved in ACSF at the desired concentrations and applied to the whole preparation. Dorsal root recordings The basal potential of the L4 or the L5 dorsal root was recorded by means of a suction electrode coupled to a Cyber-Amp amplifier (Axon Instruments, USA) set in DC mode 16

17 (further details in Rivera-Arconada and Lopez-Garcia, 2006). The spontaneous activity was quantified in terms of mean amplitude and frequency in 10 min periods. The spinal cord was challenged by three increasing concentrations of exogenous GABA applied at 10 min intervals. Depolarization in response to GABA applications were quantified as the area of depolarization (in mv*s). The sequence of three GABA boluses was applied at 45 min intervals. Three control responses were obtained prior to NS11394 application. NS11394 was applied for 15 min and responses to subsequent applications of GABA were continued for periods up to 4 h. Ventral root recordings The L5 dorsal root and the corresponding ventral root were placed in tight fitting glass suction electrodes. Electrical stimuli sufficient to activate C-fibres (200 μa and 200 μs) were applied to the dorsal root. The stimulation protocol consisted of a single stimulus followed by a 60 s rest and a train of 20 stimuli at 1 Hz. Ventral root responses to stimulation of the dorsal root were recorded using simultaneous AC-DC recording procedures as described previously (Hedo and Lopez-Garcia, 2001). Briefly, ventral root signals were split and amplified in different channels of a Cyber-Amp set to DC and AC modes, respectively. The AC channel was band-pass filtered between 300 Hz and 1500 Hz. The stimulation protocol was applied at 20 min intervals. Three basal responses were obtained prior to 15 min superfusion with NS Afterwards, stimulation was applied during drug application and up to 4 h of washout. To fully characterize the ventral root responses and the effects of the drugs, the following variables were 17

18 quantified: for DC signals from single stimuli the amplitude of the mono-synaptic reflex (in mv) and the area of depolarisation measured in a time window between 100 ms and 4 s from stimulus artefact (in mv*s); for AC signals from trains of stimuli, the number of spikes overshooting a threshold set 5 μv over the mean noise level. Under the present recording conditions, AC recordings reflect fast events such as action potentials fired by motor neurons whereas DC recordings reflect slow depolarization of motor neurons. Compounds, administration protocols and dose selection NS11394 [3 -[5-(1-hydroxy-1-methyl-ethyl)-benzoimidazol-1-yl]-biphenyl-2- carbonitrile] and zolpidem (tartrate salt) were synthesised in the Medicinal Chemistry Department, NeuroSearch A/S. Gaboxadol (chloride salt) was purchased from Sigma- Aldrich Chemie GmbH (Munich, Germany). Gabapentin was purchased from Actavis Nordic A/S (Gentofte, Denmark). Morphine hydrochloride and diazepam were purchased from Nomeco A/S (Copenhagen, Denmark). Bretazenil and flumazenil were gifts from Roche Diagnostics (Basel, Switzerland). For behavioural studies, NS11394 was dissolved in 5% Tween80/milliQ water and administered p.o. in a dosing volume of 5 ml/kg. Diazepam, bretazenil and flumazenil were dissolved in 5% cremophor (BASF, Ludwigshafen, Germany). Gabapentin was dissolved in 5% glucose/milliq water. These four drugs were administered i.p. in a dosing volume of 2 ml/kg. Gaboxadol and morphine were dissolved in 0.9% saline and zolpidem in 5% glucose/milliq water. These three drugs were administered s.c. in a 18

19 dosing volume of 1 ml/kg. All doses are expressed as either mg weight salt or mg weight free base per kg body weight, respectively. For behavioural studies we tried to ensure that over the dose range of each drug tested and using our dosing regime we had CNS penetration resulting in virtually full occupancy of GABA A receptors. Thus all four GABA A modulators showed good CNS penetration as measured by in vivo [ 3 H]flunitrazepam binding to rat cortex (>90% for diazepam (10 mg/kg), bretazenil (5 mg/kg) and NS11394 (5 mg/kg) and ~70% for zolpidem (20 mg/kg); unpublished data; see also Mirza and Nielsen, 2006; Mirza et al., 2008). Although zolpidem appeared at the bottom end of the desired range this is slightly misleading due to its affinity selectivity for GABA A -α1 receptors (Sanger and Benavides, 1993). Despite this lower level of in-vivo receptor occupancy with zolpidem, it is very clear from the data described in the Results section that major CNS mediated effects are seen at doses between 1-10 mg/kg, and it was therefore not meaningful to dose higher. With respects to gaboxadol, pharmacokinetic data in the rat show clear CNS penetration and CNS related effects at the doses utilised here (Cremers and Ebert, 2007). For electrophysiology studies, NS11394 was diluted in DMSO at 10 mm and stored at - 20ºC. NS11394 was diluted down to 1 μm in ACSF just prior to use and superfused to the entire preparation for 15 min periods to ensure a complete tissue equilibration. Application of 0.01% DMSO alone did not have any observable effect on spinal reflexes (n = 2). GABA was dissolved in ultrapure water at 10 mm stored at -20ºC, 19

20 diluted down to its final concentration in ACSF just prior to use where it was delivered to the preparation as a 1 min bolus. Data analysis Comparison of drug effects between injured and non-injured rats across the range of behavioural tests employed was made via conversion of raw data to either % of vehicle, baseline or maximal possible effect values. Analysis of data was performed using Sigmastat 2.03 (SPSS Inc., Chicago, ILL.). All data are presented as mean ± SEM. Unless stated otherwise, ANOVA was used to analyze the overall effects of treatments. When the F value was significant this was followed by Bonferroni s all pairwise comparison. p<0.05 was considered to be statistically significant. For electrophysiology studies, statistical analysis was performed on raw data using GraphPad Prism (v3.02). Repeated measures ANOVA was used to analyze responses to trains of stimuli. Results are expressed as mean ± SEM. p<0.05 was considered to be statistically significant. 20

21 Results Explorative motility and Rotarod test To determine if the analgesic properties of GABA A receptor modulators described below were specific and not confounded by any extraneous properties, we evaluated their central nervous system side-effect profile in studies of explorative motility and activity-induced motor function in non-injured rats (Figure 2A). Rats administered vehicle displayed typical explorative behaviour of a novel environment throughout a 30 min test (range of 5 experiments 52 ± ± 3 m). Although administration of NS11394 (3-120 mg/kg), diazepam (1-5 mg/kg), zolpidem (1-10 mg/kg), gaboxadol (1-6 mg/kg) and bretazenil (10-60 mg/kg) all attenuated explorative motility behaviour (F[5,55]=2.474, p<0.05; F[3,27]=29.277, p<0.001; F[3,27]=35.822, p<0.001 and F[3,23]=31.608, p<0.001; F[3,27]=3.69, p<0.05, respectively) it was clear from further analysis that these compounds differed in their potency and magnitude of effect (Figure 2A), with NS11394 and bretazenil having minor effects relative to the other drugs. In the rotarod test (Figure 2B), NS11394 (1-120 mg/kg) and bretazenil ( mg/kg) engendered a modest, but nevertheless significant ataxia (F[6,71]=3.303, p<0.01 and F[3,28]=4.193, p<0.05). As in the explorative motility experiment, only exceptionally high doses of NS11394 and bretazenil significantly attenuated motor function compared with corresponding vehicle responses (-1.9 ± 1.1 s and -15 ± 8.5 s; both p<0.05, Figure 2B). In contrast, the modulators diazepam (1-5 mg/kg) and zolpidem (1-10 mg/kg) as well as the agonist gaboxadol (1-6 mg/kg) all produced marked dose-dependent deficits in motor function (F[3,31]=48.084, p<0.001; F[3,31]=89.960, p<0.001; F[3,30]=28.138, p<0.001). Finally, as expected over the dose range studied ( mg/kg), the 21

22 antiepileptic drug gabapentin, routinely used in the clinical treatment of neuropathic pain, dose-dependently impaired motor function compared with vehicle treatment (F[3,31]= p<0.001). These findings are wholly consistent with the literature and confirm that partial allosteric modulators (NS11394, bretazenil) contrary to full allosteric modulators (diazepam, zolpidem) and GABA site agonists (gaboxadol) display a very benign CNS side effect profile (Haefely et al., 1990; Ebert et al., 2006). However, it is important to emphasize that in our experience and observations here, zolpidem s influence on motoric integrity in the rat is more profound than that seen with diazepam, despite seemingly similar impacts on explorative motility and rotarod ataxia (Figure 2). Hot plate test To test for possible analgesic actions on acute nociceptive processing, GABA A receptor modulators were administered to normal, uninjured rats in the hot plate test (Figure 3). Vehicle responses were -2.0 ± 1.0, -2.7 ± 1.3, -3.8 ± 1.1, 0.8 ± 0.5, -0.8 ± 0.5 and -3.0 ± 1.3 s for NS11394, diazepam, zolpidem, gaboxadol, bretazenil, and morphine experiments, respectively. One way ANOVA failed to reveal any effect of NS11394 (3-30 mg/kg), diazepam (2-20 mg/kg), zolpidem (1-10 mg/kg) or bretazenil (3-30 mg/kg) treatment on the latency to respond to noxious thermal stimulation of either hindpaw (F[3,31]=2.616, p=0.071; F[3,31]=0.639, p=0.596; F[3,31]=2.870, p=0.054; F[3,30]=2.581, p=0.074). In contrast, gaboxadol (1-10 mg/kg) exhibited a modest but significant analgesic-like profile in this test (F[3,31]=7.575, p<0.001), whilst the μ- 22

23 opioid receptor agonist morphine (1-10 mg/kg) was as expected fully analgesic under the conditions tested (F[3,31]=79.978, p<0.001). Formalin test and capsaicin-induced sensitization Injection of formalin into the rat hindpaw selectively activates TRPA1 containing C- fibres and initiates spontaneous nociceptive behaviours consisting of either flinching, and licking and/or biting of the injected paw (McNamara et al., 2007). The first phase can be attributed to direct chemical stimulation of nociceptors, interphase to activation of noxious inhibitory controls and the second phase to peripheral inflammatory processes and subsequent sensitization of nociceptive spinal neurones (Coderre et al., 1993; Henry et al., 1999). Injection of capsaicin into the rat hindpaw selectively activates and sensitizes TRPV1 containing C-fibre primary afferents and manifests as flinching, and licking and/or biting of the injected paw. NS11394 (3-30 mg/kg) significantly reduced flinching behaviour during interphase (F[3,30]=4.139, p<0.05) and the second phase (F[3,30]=11.033, p<0.001) of the formalin test compared with vehicle treatment indicative of a selective effect on injuryinduced nociceptive transmission (Figure 4A and Table 1). Similarly, diazepam (1-10 mg/kg) and zolpidem (1-10 mg/kg) appeared to selectively attenuate second phase flinching behaviour (F[3,31]=6.880, p<0.01 and F[3,29]=7.234, p<0.01, respectively; Table 1). Gaboxadol (3-10 mg/kg) was the only compound tested that reduced flinching throughout all three phases of the test (F[3,31]=14.797, p<0.001; F[3,31]=20.484, p<0.001; F[3,31]=45.495, p<0.001 for first phase, interphase and second phase, respectively), indicative of a non-selective effect on pathological nociceptive 23

24 transmission in this model. Interestingly, formalin-induced flinching was completely unaffected by administration of bretazenil (3-30 mg/kg). Finally, as expected gabapentin ( mg/kg) attenuated second phase flinching behaviour (F[3,31]=5.320, p<0.01). To verify that the antinociceptive actions of NS11394 were mediated selectively via GABA A receptors, the benzodiazepine site antagonist flumazenil was administered in combination with NS Administration of NS11394 (10 mg/kg) alone produced a 33% reduction in flinching during the second phase of the test (p<0.05 vs vehicle; Figure 4B), and this effect was blocked by co-administration of flumazenil (30 mg/kg). Intriguingly, although flumazenil alone had no effect on second phase flinching, an increase in the number of flinches during interphase compared with vehicle was observed (p<0.05). Therefore we re-investigated various doses of flumazenil (3-30 mg/kg, s.c., n=8 rats per group) in a separate study and found only a small and nonsignificant (20 ± 17 %) increase in flinching during interphase at 30 mg/kg (data not shown), indicating that this was not a robust finding. Compared with vehicle treatment (156 ± 22 total flinches) the hindpaw flinching observed within 5 min after capsaicin injection was unaffected by NS11394 ( mg/kg, Figure 5). Over the following 25 min which would be expected to coincide with a more persistent and selective activation of TRPV1 containing C-fibres, vehicle-treated rats continued to exhibit a robust flinching response (400 ± 44 total flinches). During this period, NS11394 treatment significantly reduced flinching behaviour (F[5,57]=8.701, p<0.001, Figure 5). Administration of diazepam (1-10 mg/kg), zolpidem (1-6 mg/kg) and gaboxadol (3-10 mg/kg) also dose-dependently attenuated flinching from 6-30 min (F[3,31]=9.294, p<0.001, F[3,31]=4.017, p<0.05, F[3,27]=34.448, p<0.001, respectively; Figure 5B) after capsaicin injection. However, 24

25 in addition diazepam and gaboxadol both dose-dependently attenuated flinching from 0-5 min (F[3,31]=6.539, p<0.01 and F[3,27]=10.626, p<0.001, respectively, data not shown). Capsaicin-induced flinching was completely unaffected by administration of bretazenil (10-60 mg/kg). Complete Freunds adjuvant-induced inflammatory nociception Peripheral inflammation increases inhibitory synaptic transmission mediated by GABA A receptors within the spinal dorsal horn (Poisbeau et al., 2005). At the behavioural level, gait deficits associated with established monoarthritis are completely prevented by continuous administration of muscimol (Simjee et al., 2004). Thus, we compared antinociceptive actions of NS11394 and diazepam relative to morphine in the CFA model of inflammatory pain. Twenty-four hours after injection of CFA into the hindpaw a marked alteration in hindpaw weight bearing indicative of spontaneous nonevoked pain was observed compared with weight bearing prior to injection (35 ± 1.9 g vs -0.1 ± 3.2 g, n=95, p<0.001, Students t-test). Injection of NS11394 (1-10 mg/kg) markedly attenuated the deficit in hindpaw weight bearing (F[4,61]=7.569, p<0.001) in CFA rats (Figure 6A). In contrast, the tail flick response in the same CFA rats was completely unaffected by NS11394 (F[3,30]=0.215, p=0.885). One way ANOVA also revealed a significant antinociceptive effect of diazepam ( mg/kg) treatment against hindpaw weight bearing deficits in CFAtreated rats (F[4,63]=7.672, p<0.001). However, in contrast to NS11394, diazepam also significantly affected the tail flick response in CFA treated rats (F[3,31]=6.740, p<0.01, Figure 6B). As expected, morphine (3-10 mg/kg) significantly attenuated weight 25

26 bearing deficits (F[4,63]=21.896, p<0.001) and increased tail flick latency responses in CFA rats (F[3,31]=91.094, p<0.001; Figure 6C). Peripheral nerve injury SNI and CCI To evaluate drug effects upon the reflex sensory component of the nociceptive response associated with evoked neuropathic pain behaviours, GABA A receptor modulators/agonists and gabapentin were tested for antiallodynic effects in the CCI model of peripheral nerve injury. NS11394 and gabapentin were also assessed for antiallodynic and antihyperalgesic effects in the SNI model of neuropathic pain. Following surgery, both CCI (n=103) and SNI (n=34) rats developed behavioural signs of mechanical allodynia (observed as a decrease in the paw withdrawal threshold in response to von Frey hair stimulation) of the ipsilateral hindpaw (1.1 ± 0.1 g and 0.5 ± 0 g, respectively), compared to pre-surgery levels that typically measured 13.5 g. Mechanical hyperalgesia (observed as an increase in the paw withdrawal duration in response to pin prick stimulation) of the ipsilateral hindpaw was also observed in both nerve injury models (15 ± 0 s and 14 ± 0.3 for CCI and SNI rats, respectively) compared to pre-surgery levels that were <0.5 s. In CCI rats, mechanical allodynia was dose-dependently reversed (F[3,23]=5.021, p<0.01) by administration of NS11394 (5-30 mg/kg, Figure 7A). Similarly, gaboxadol (3-10 mg/kg) dose-dependently reversed hindpaw mechanical allodynia in CCI rats (F[3,23]=20.244, p<0.001). Although diazepam also reversed mechanical allodynia (F[3,23]=20.732, p<0.001) in CCI rats, the attenuation in hindpaw hypersensitivity was significant only for the highest dose of diazepam tested (p<0.001; 10 mg/kg vs vehicle). 26

27 By contrast, both zolpidem (3-10 mg/kg) and bretazenil (10-60 mg/kg) were completely ineffective at reversing hindpaw mechanical allodynia in CCI rats. In SNI rats, two way repeated measures ANOVA revealed a clear interaction between NS11394 (3-30 mg/kg) treatment with respect to time on hindpaw withdrawal threshold (F[9,95]=12.537, p<0.001, Figure 7B). In terms of comparative antiallodynic efficacy, the magnitude of the reversal obtained with 30 mg/kg NS11394 in SNI and CCI rats (89 and 71%, respectively) was equivalent to that obtained with the antiepileptic drug gabapentin (200 mg/kg) in the same injury models (94% and 78% for SNI and CCI, respectively, Figure 7C). Similarly, both NS11394 and gabapentin dose-dependently reversed hindpaw mechanical hyperalgesia in SNI rats (F[3,23]=11.745, p<0.001 and F[3,23]=17.804, p<0.001). Again, the magnitude of reversal obtained was comparable in SNI rats administered either 30 mg/kg NS11394 or 200 mg/kg gabapentin (86 vs 90%, respectively; data not shown). In vitro electrophysiology In the light of a spinal action of diazepam (Siarey et al., 1992; Siarey et al., 1994) we performed electrophysiological experiments on the isolated spinal cord to learn about possible mechanisms of action of NS11934 in the spinal circuits that mediate withdrawal reflexes. To this end we first checked the effects of NS11394 on native spinal cord GABA A receptors localized to the central terminals of primary afferents. NS11394 (n=3) applied at 1 μm produced a long-lasting potentiation of GABA-induced primary afferent depolarization for all three concentrations of exogenously applied GABA (Figure 8A). The mean integrated area of depolarization produced by GABA 27

28 applied at 30 μm increased from 19 ± 1.2 mv*s in control ACSF to 84 ± 2.9 mv*s after superfusion of NS11394, corresponding to a 454 ± 44% increase. In addition, application of NS11394 produced an increase in amplitude (133 ± 7.5% of control) and frequency (167 ± 12%) of spontaneous dorsal root potentials in all three preparations. To evaluate the effects of NS11394 on the overall spinal integration of afferent signals we recorded dorsal root-ventral root reflexes (DR-VRr) in the absence and presence of NS The analysis of responses to single stimuli showed that the monosynaptic reflex, mediated by the activation of thin myelinated fibres, was unaltered by NS11394 (n=3). In contrast, longer latency components of the DR-VRr, mediated by activation of nociceptive afferents, were strongly and systematically depressed by NS11394 (Figures 8B and C). The integrated area of the response decreased to 60 ± 1.5 % of control after application of NS This reduction was due to a fall in the amplitude of the depolarization most clearly seen at latencies exceeding 300 ms from stimulus artefact (Figure 8C). Repetitive stimulation of the dorsal root produced a typical wind-up of action potentials which depends on C-fibre activation. The slope of the action potential wind-up was eliminated by superfusion with NS11394 (n=3) and the total count of action potentials to this repetitive stimuli fell to 9 ± 4% of control (Figures 8D and E). 28

29 Discussion NS11394 is a positive allosteric modulator at GABA A receptors with a subtypeselectivity profile that translates to marked anxiolytic properties at doses devoid of motoric impairment in rodents (Mirza et al., 2008, summarised in Figure 1). Recent studies in gene targeted animals suggest a pivotal role for GABA A -α2/gaba A -3 receptors in normalising spinal disinhibition after injury. Therefore we have tested orally administered NS11394 against a range of injury-induced nociceptive behaviours in rat models associated with central sensitization. Persistent and inflammatory nociception In the rat formalin test spinal administration of GABA A receptor agonists and modulators have been reported to produce conflicting effects on nociceptive behaviours (Dirig and Yaksh, 1995; Kaneko and Hammond, 1997). In our experiments, NS11394 had no effect on first phase flinching after injection of formalin or capsaicin, consistent with its lack of effect on acute nociceptive behaviours (see below). Crucially, NS11394 reduced second phase formalin- and capsaicin- induced flinching. The efficacy of NS11394 in the formalin test was equivalent to that of gabapentin. Given that persistent nociceptive transmission mediated by TRPA1- (formalin) and TRPV1- (capsaicin) containing C-fibre afferents was similarly attenuated by NS11394, this indicates a common site of integrative antinociceptive action, perhaps on postsynaptically located GABA A receptor-containing neurones within the spinal dorsal horn. However, supraspinal sites of action cannot be excluded, and indeed there is a close correlation between the minimum effective dose of NS11394 for second phase antinociception (1-3 mg/kg) with the ED 50 to displace cortical [ 3 H]flunitrazepam binding (1.3 mg/kg) in vivo 29

30 (Mirza et al., 2008). Regardless, we can ascribe NS11394-mediated antinociception in the formalin test to binding to GABA A receptors in vivo since flumazenil reversed the effect of NS Diazepam, zolpidem and gaboxadol, but not bretazenil, also had antinociceptive effects in the formalin and capsaicin models with efficacy essentially on a par with NS11394, albeit at doses associated with motoric impairments. In non-injured rats, spinal administration of the GABA A agonist isoguvacine produces a bicuculline-reversible increase in acute nociceptive threshold. To investigate if NS11394 would selectively attenuate behavioural sequelae associated with injuryinduced central sensitization, its influence on spontaneous nociceptive behaviours in the CFA model of inflammatory pain was compared with its actions on acute nociceptive reflexes. Hindpaw weight bearing deficits induced by CFA were exquisitely sensitive to treatment with NS11394, which displayed full efficacy at all doses tested. However, NS11394 had no effect on tail flick latency in CFA rats and did not affect hot plate latency responses in non-injured rats, indicating no influence on acute nociceptive responses. By contrast, diazepam increased acute nociceptive threshold in CFA-treated animals in addition to altering hindpaw weight bearing deficits. Neuropathic injury Whereas spinal administration of bicuculline induces mechanical allodynia in normal rats (Malan et al., 2002), spinal administration of muscimol reverses mechanical allodynia in the spinal nerve ligation model of neuropathic pain (Hwang and Yaksh, 1997; Malan et al., 2002). Importantly, these effects on allodynia occur at spinal doses that minimally affect motor function, implying that robust and selective antinociception 30

31 is possible by specifically targeting GABA A receptors localised within central pain pathways (Hwang and Yaksh, 1997; Knabl et al., 2008). Our data support this hypothesis. NS11394 reversed mechanical allodynia in SNI and CCI rats, to a level comparable with gabapentin which is routinely used in the clinical treatment of neuropathic pain. Whereas similar antiallodynic efficacy was achieved with diazepam and gaboxadol in CCI rats, effective doses of both drugs were associated with overt sedation and ataxia (Table 2). By contrast, bretazenil was inactive in the CCI model, and indeed all models of pain. Interestingly, despite doses of zolpidem tested in CCI rats leading to clear motoric impairment it was inactive in this model of neuropathic pain. What are the pharmacological determinants of NS11394 s antinociceptive actions? Diazepam has full efficacy at α1, α2, α3 and α5 containing GABA A receptors and is effective across the range of pain models utilised here. Since effective doses of diazepam in these models were associated with CNS related motor side effects, one might reasonably speculate that the antinociceptive effects of diazepam are secondary to side effects. However, we do not believe the explanation is so straightforward, since broad spectrum antinociception was not seen with zolpidem which severely affected motor function. From our data with NS11394, we conclude that robust antinociception with minimal influence on basic motor function can be obtained with only partial allosteric modulation of GABA A receptors. However, one key finding in our dataset is the complete lack of efficacy of bretazenil across all models of pain, despite this compound being very potent ( 0.3 mg/kg) and efficacious in animal models of anxiety 31

32 (Haefely et al., 1990). A logical conclusion from this is that bretazenil simply has insufficient efficacy at GABA A receptors (Sieghart, 1995) to produce antinociception in animal models, which implies that efficacy requirements for anti-anxiety and antinociceptive effects differ. Another interesting finding is the inconsistent efficacy of zolpidem across the range of pain models. Since zolpidem has ~10-fold higher affinity for GABA A -α1 compared to GABA A -α2 and GABA A -α3 receptors (Sieghart, 1995) this indicates that GABA A -α1 receptors are not involved in antinociception, consistent with NS11394 and L838,417 data (Knabl et al., 2008). However, zolpidem is still a full efficacy modulator of GABA A -α2 and GABA A -α3 receptors so why does it not work in CCI? We believe that although zolpidem might have sufficient activity at GABA A -α2 and GABA A -α3 receptors to engender antinociceptive activity, this is masked in some models by severe motor side-effects. An alternative and intriguing explanation is that there is a role for GABA A -α5 receptors. Thus, whereas zolpidem s efficacy at GABA A -α2 and GABA A - α3 containing receptors is sufficient to mediate antinociceptive actions in formalin/capsaicin models, its poor affinity for GABA A -α5 receptors (Sieghart, 1995) precludes efficacy in CCI rats. Although speculative this explanation is consistent with our data as diazepam, NS11394 and L838,417 all have efficacy at GABA A -α5 receptors. GABA A -α5 containing receptors are distributed through numerous spinal lamina, and α5 gene expression is modulated in the DRG and dorsal horn in neuropathic pain models (Xiao et al. 2002; Yang et al., 2004). 32

33 Limited studies with α4 knockout mice implicate GABA A -α4 receptors in gaboxadols antinociceptive properties (Chandra et al., 2006). However, this receptor is unlikely to be relevant for NS11394 and diazepam induced antinociception since both bind poorly to GABA A -α4 receptors (Mirza et al., 2008). Indeed, although at the doses used here it has been suggested that gaboxadol has a higher activity level at extrasynaptic GABA A - α4, α5 and α6 receptors (Ebert et al., 2006; Cremers and Ebert, 2007), we would urge future studies to investigate the possibilities that gaboxadol s antiociceptive actions might be mediated, (i) via GABA A -α3 and/or GABA A -α2 receptors; or (ii) possibly via GABA A -α5 receptors, in reference to the discussion on zolpidem above. Possible mechanisms of action of NS11934 Our exploratory spinal cord electrophysiology studies confirm that NS11394 enhances the inhibitory effect of native GABA A receptors and profoundly depresses the activity of spinal circuits which mediate nociceptive withdrawal reflexes. Primary afferent depolarization (PAD), mediated predominantly by GABA A receptors, constitutes a classical mechanism for pre-synaptic inhibition (Rudomin and Schmidt, 1999). Here we show that NS11394 potentiates PAD induced by exogenous GABA and spontaneous activity recorded from primary afferents - both phenomena known to be sensitive to picrotoxin (Rivera-Arconada and Lopez-Garcia, 2006). NS11394 s actions are specific since it did not alter the monosynaptic reflex, which in the rat is sensitive to baclofen but not to bicuculline (Akagi et al., 1987). These observations indicate that the native spinal GABA-ergic system is modulated by NS11394 as predicted from studies performed on recombinant GABA A receptors (Mirza et al., 2008). Furthermore, these results indicate that NS11394 modulates sensory processing in the dorsal horn. 33